Frontal dysconnectivity in 22q11.2 deletion syndrome: an atlas-based functional connectivity analysis

Chromosome 22q11.2 deletion syndrome (22q11DS) is caused by a microdeletion of approximately 50 genes on one copy of the q11.2 band of chromosome 22. Youth with the syndrome typically present with physical anomalies, cognitive impairments, and behavioral disorders [1, 2]. During adolescence and young adulthood, approximately 30–40% of individuals with 22q11DS develop a psychotic illness, usually schizophrenia [3–5]. This represents a significant increase over the risk for schizophrenia in the general population [6]. The neurobiological mechanisms underlying this increased risk for schizophrenia in individuals with 22q11DS are not well-understood.

Converging evidence supports the notion that idiopathic (non-syndromal) schizophrenia is a disorder of functional and structural dysconnectivity [7–11]. Studies of functional connectivity point to a preponderance of anomalies in frontal–temporal connectivity [12, 13], although frontal–parietal and frontal–occipital connections have also been implicated [14, 15]. Moreover, abnormalities have been observed in several large-scale, functional networks, including the default mode network, the salience network and the central executive network [16–18].

Although studies examining functional dysconnectivity in 22q11DS are much fewer in number, the findings are consistent with studies of idiopathic schizophrenia [19]. Results of these studies indicate anomalous connectivity in frontal lobe connections [20] and parieto–occipital connections [20–22]. Decreases in functional connectivity have also been observed, in partially overlapping samples, in the default mode [23–26], salience [24] and frontal–parietal networks [22, 24]. In a modularity analysis of overall functional network organization, Scariati and colleagues [27] observed increased modular segregation across superior parietal, frontal and inferior temporal lobes in individuals with 22q11DS. Associations between anomalous functional connectivity in 22q11DS and increased symptoms of psychosis have been observed in most [20, 22, 24], but not all studies [25].

To our knowledge, two studies by Scariati and colleagues [20, 27] have conducted a functional connectivity analysis of atlas-based, ROI-to-ROI structural connections in 22q11DS. Scariati and colleagues first reported widespread functional connectivity in individuals with 22q11DS, primarily affecting frontal and temporal lobe regions. In a more recent study [27], they focused on age differences by examining connectivity in a sample of 9–30 year-old individuals with 22q11DS that were divided into two age groups (groups split at 18 years old) for subanalyses. In both age groups, alterations of modular communities were found to affect the anterior cingulate cortex and parieto-occipital processing regions. However, in adults with 22q11DS, they observed nontypical modularity partition of the dorsolateral prefrontal cortex.

Here, we conduct an atlas-based functional connectivity analysis of ROI-to-ROI connections in individuals with 22q11DS who are specifically between the ages of 18 and 24 years, a time-frame that poses the greatest risk for developing psychotic illness. In this ROI-to-ROI based approach, we sought to assess connectivity patterns by matching an anatomical atlas to each subject’s own fMRI space. The methodological advantage of this approach is that data were not normalized to a standard template, thus obviating potentially problematic effects of warping the brain. Conceptually, a subject-specific, atlas-based approach can yield additional data about the functional architecture and organization of the brain [28, 29]. Moreover, the use of atlas-based ROIs provides a common framework to increase reproducibility across studies, and can be incorporated for use in multimodal studies. In order to implement this approach, we applied the functional connectivity toolbox, CONN [28–30], which has shown a high degree of interscan reliability [28] and has demonstrated disease-relevant functional connections between anatomically defined regions of the brain [30]. We hypothesized that ROI-to-ROI connectivity between sublobar frontal–parietal gyri, and frontal–temporal gyri would be anomalous in individuals with 22q11DS relative to controls, and that aberrant connectivity would be associated with symptoms of psychosis.

Using a seed-based connectivity analysis of 2278 ROI-to-ROI connections, we observed both hyper- and hypo-connectivity in frontal–frontal gyri, frontal–parietal gyri, frontal–occipital gyri, parietal–temporal gyri and occipital–occipital gyri in young adults with 22q11DS relative to controls. Notable findings included (1) increased functional connectivity between frontal (superior frontal, caudal middle frontal and pars orbitalis) gyri and the precuneus, and (2) increased functional connectivity between posterior cingulate gyrus and both superior frontal and precentral gyri. Anomalies in frontal–parietal and occipital–occipital gyral connectivity were significantly associated with positive symptoms of psychosis.

The precuneus, caudal middle frontal and pars orbitalis (i.e., medial inferior frontal) regions constitute part of the default mode network (DMN), which as noted above, is reported to be anomalous in both schizophrenia and 22q11DS. Studies have demonstrated that the DMN is active not only during rest but also during activities involving self-referential [41] and social-interpersonal processing [42]. Evidence suggests that the DMN may be involved in auditory hallucinations in individuals with schizophrenia [43–45], although other networks have been implicated as well [46, 47]. In individuals with 22q11DS, the DMN has been associated with prodromal symptoms [21], sustained attention [21] and reciprocal social behaviors [23]. It is not clear why we observed increased functional connectivity between these DMN regions, while several other studies [23–26] of 22q11DS have observed decreased functional connectivity between these regions. This may be attributable, in part, to our implementation of measurements within each subject’s native brain space. In light of the anatomic differences that have been reported in brains of individuals with 22q11Ds, retaining each subject’s native brain space may have produced results that are not totally (anatomically) comparable to studies in which brains are warped to a standard template. Moreover, potential differences in sample characteristics (e.g. IQ levels; medication usage) between studies may also be contributing to differences in the direction of these results (see review by Scariati and colleagues [19]). Additional insight into why our finding of increased functional connectivity in the DMN differs from several (but not all [21, 22]) studies of 22q11DS is suggested by the results of two previously-published papers [22, 24] that included samples that overlapped with the sample of the current. In our two previously-published papers, we pooled samples from two research sites, and applied Independent Components Analyses to the pooled data. However, preprocessing methods differed somewhat between the two papers. In the first paper, by Mattiaccio and colleagues [22], for which data were preprocessed and analyzed at our site, increased functional connectivity in the DMN was observed. In the second paper, by Schreiner and colleagues [24], the data were preprocessed and analyzed by our collaborating site, and decreases in functional connectivity in the DMN were observed. Interestingly, our respective sites’ preprocessing methods differed in motion correction and noise reduction strategies, potentially accounting for the discrepancies in results. This supports the notion that differences in image processing methods and in sample characteristics may be contributing to between-study differences in results.

The posterior cingulate gyrus (PCG) is also part of the default mode network, and we found anomalies in connectivity between PCG and superior frontal and precentral gyri. The extent to which PCG—superior frontal connections in our study reflects the DMN is not completely clear, since we utilized a predefined, atlas-based approach that maps onto regions that subsume, but are not synonymous with the DMN. Nonetheless, primate (and more recently, human imaging) studies indicate that the PCG has strong, reciprocal connections to the dorsolateral prefrontal cortex (DLPFC) [48–50], which overlaps with the superior frontal region included in the Desikan-Killany atlas. It has been suggested that PCG-DLPFC connections may be part of both the dorsal attention network and the frontal-parietal control network [51] both of which contribute to efficient cognitive function. Functional connectivity of the PCG and the superior aspect of the DLPFC has been linked to goal-directed thought processes [52], suggesting that this reciprocal connection may subserve executive planning [53, 54] and cognitive control [53, 55], both of which are impaired in individuals with 22q11DS [56–59]. Moreover, these functional brain networks have been shown to be impaired in schizophrenia [14, 60, 61] and 22q11DS [22, 24, 62].

Of the 16 ROI-to-ROI connections that significantly differentiated individuals with 22q11DS from controls, 13 (81%) of them included at least one ROI in the frontal lobe. These findings are consistent with other functional connectivity studies of both idiopathic schizophrenia [7, 12, 13, 63] and 22q11DS [20, 23] and suggest that both short-range and long-range connectivity of the frontal lobe is anomalous in individuals with this syndrome. To the extent that the frontal lobe subserves a myriad of cognitive and social-affective functions, functional dysconnectivity of networks that include the frontal lobe could underlie many of the cognitive and psychiatric impairments that are associated with 22q11DS [20, 23]. For example, in addition to schizophrenia, frontal dysconnectivity has been implicated in both autism spectrum disorders and in ADHD, both of which are elevated in 22q11DS [5, 57, 64–68].

In our sample, positive prodromal symptoms of psychosis were associated with increased connectivity between the superior frontal gyrus and the precuneus, and with decreased connectivity between the right and left pericalcarine gyri of the occipital lobe, and between pericalcarine and postcentral gyri. As noted above, the precuneus and aspects of the superior frontal gyrus are included in the DMN, which previous studies of 22q11DS have associated prodromal symptoms as well [21]. Associations between parietal–occipital and occipital–occipital functional connections and prodromal symptoms of psychosis have not been reported. However, anatomic connections between parietal and occipital lobes, via the superior longitudinal fasciculus (SLF), have been reported to be aberrant in 22q11DS [69–72]. Moreover, in an overlapping sample, our group [73] recently reported associations between anatomic anomalies in the SLF and prodromal symptoms.

When we divided the group of individuals with 22q11DS into prodromal and nonprodromal subgroups, we observed a significant difference in connectivity between the left inferior temporal and right pericalcarine gyri. Interestingly, we recently reported (in the same patient sample) significant associations between white matter microstructural anomalies in the temporal-occipital aspect of the inferior longitudinal fasciculus and symptoms of psychosis [74]. Temporal-occipital alterations in functional connectivity have also been reported in patients experiencing their first episode of psychosis [75], further supporting the validity of these observations.

Our study utilized an atlas-based approach to investigate functional connectivity in 22q11DS, which permitted us to examine, within each individual’s own fMRI space, more than 2000 functional connections throughout the cortex. A potential limitation to our method is that the acquisition time of 5 min that we used to acquire our fMRI data, while minimally acceptable for an fcMRI study, may not be optimal in order to minimize the effects of noise and ensure the detection of small correlations that might otherwise go unobserved [76]. A second potential limitation is that the connections we examined do not necessarily map specifically onto the networks that are traditionally examined in resting state fcMRI studies, thus limiting comparisons to other studies to some extent, and rendering conclusions regarding these comparisons somewhat speculative. Nonetheless, our results concur generally with previous studies that have observed DMN anomalies in 22q11DS and associations between DMN anomalies and prodromal symptoms of psychosis. However, we observed increased functional connectivity in DMN regions, in contrast to several previous studies that have observed reduced connectivity. As noted above, this may be due in part to the potential impact of current medication usage in our sample, and to study differences in image preprocessing. In addition, it should be noted that when we removed the subset of controls with ADHD and anxiety, study group differences in the connections between the PCG and both the superior frontal and precentral gyri did not survive correction for multiple comparisons. This may suggest that the presence of psychiatric disorders in our sample may be influencing our observation of study group differences in connectivity between PCG and other frontal-based regions; however, the removal of the control subgroup also reduced power to detect differences. Accordingly, future studies would benefit from larger samples to elucidate the potential interplay between the presence of psychiatric disorders in 22q11DS and functional connectivity. To the extent that sampling and image preprocessing differences account for discrepancies across studies, it would be useful, in general, to apply different preprocessing methods to identical samples in order to elucidate the extent to which these methods account for differences in results of functional connectivity studies. Within the area of neurofunction in 22q11DS, future studies should examine the associations between functional and structural connectivity in 22q11DS, in order to elucidate the extent to which neuroanatomic structure underlies functional anomalies and leads to the psychiatric impairments for which individuals with this disorder are at great risk.